<<

ARTICLE

DOI: 10.1038/s41467-018-05052-4 OPEN Selective ensembles in supported palladium sulfide nanoparticles for semi-

Davide Albani1, Masoud Shahrokhi2, Zupeng Chen1, Sharon Mitchell1, Roland Hauert 3, Núria López 2 & Javier Pérez-Ramírez1

Ensemble control has been intensively pursued for decades to identify sustainable alter-

natives to the Lindlar catalyst (PdPb/CaCO3) applied for the partial hydrogenation of

1234567890():,; in industrial . Although the geometric and electronic requirements are known, a literature survey illustrates the difficulty of transferring this knowledge into an efficient and robust catalyst. Here, we report a simple treatment of palladium nanoparticles supported on graphitic carbon nitride with aqueous sodium sulfide, which directs the for-

mation of a nanostructured Pd3S phase with controlled crystallographic orientation, exhibiting unparalleled performance in the semi-hydrogenation of alkynes in the liquid phase. The exceptional behavior is linked to the multifunctional role of . Apart from defining a structure integrating spatially-isolated palladium trimers, the active ensembles, the modifier imparts a mechanism and weak binding of the organic intermediates. Similar

metal trimers are also identified in Pd4S, evidencing the pervasiveness of these selective ensembles in supported palladium sulfides.

1 Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zürich, Vladimir-Prelog-Weg 1, 8093 Zürich, Switzerland. 2 Institute of Chemical Research of Catalonia (ICIQ), and The Barcelona Institute of Technology Av., Països Catalans 16, 43007 Tarragona, Spain. 3 EMPA, Swiss Federal Laboratories for Materials Science and Technology, Uberlandstrasse129, 8600 Dübendorf, Switzerland. These authors contributed equally: Davide Albani, Masoud Shahrokhi. Correspondence and requests for materials should be addressed to N.Lóp. (email: [email protected]) or to J.Pér-Rír. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4

he ultimate mission of researchers working in hetero- Liquid phase Gas phase geneous isthe 2010 identification 1990 of active 1970 ensembles 1950 1930 1890 T fi [1] with a high density to enable ef cient turnover while Pt black avoiding undesired intermolecular interactions, and exhibiting long-term stability to ensure a well-defined performance. Ni, Cu[2] Nevertheless, the fulfillment of all these requirements in one Pd catalytic material has yet to be accomplished, even for well- 1910 established reactions like heterogeneously-catalyzed hydrogena- tions studied since the 1880’s by Wilde, Sabatier, and 1,2 Senderens . Pd colloids[6] The chemo- and stereoselective hydrogenation of functiona- lized alkynes is a key catalytic transformation applied industrially PdPb at different scales, ranging from the gas-phase multi-ton pur- Ni[7] [8] ification of olefin streams for polymer manufacture3 to the Pd/diatomite Ni Raney[9`] smaller three-phase batch preparation of high-added-value fine chemicals4,5. Examination of the rich scientific literature emphasizes the wide dominance of palladium-based catalysts, due Alkali metals[10] to the ability to dissociate , and the importance of [11] PdPb/CaCO3 preventing the undesired hydrogenation of competing functional Ag PdAg/Al O [12] groups through electronic poisoning with suitable modifiers 2 3 – Pd/BaSO + KOH[13] (Fig. 1)6 35. Similarly, oligomerization paths (C–C couplings) 4 must be suppressed by the strict control over the ensemble size. [14] – Pd/CaCO3 + SnCl2 Thus, the synthesis of trimeric ensembles36 39, and the presence of electronic modifiers reducing the hydrogen uptake of Pd and [15] the formation of subsurface hydrides, are key parameters for PdPb/CaCO3+ Bn2S achieving high selectivity both in gas- and three-phase operations. Pd-HHDMA Since 1952, the manufacture of building blocks for pharmaceu- ticals, vitamins, agrochemicals, and fragrances via alkyne semi- Pd/ZnO[16] hydrogenation4,40,41 has been conducted over the well-known lead- PdAu/TiO [18] poisoned palladium (5 wt.%) supported on calcium carbonate (Lin- c-Pd/C[17] 2 dlar catalyst), despite drawbacks of poor metal utilization and toxicity42,43. However, the recent disruptive legislations restricting the CuNiFe[19] CeO2 use of hazardous compounds have steered the innovation and PdGa[20] PdZn/ZnO[21] creativity of scientists towards the development of sustainable alter- Pd/TiO + Ph S[22] [25] 2 2 natives. This has enabled the incorporation of new families of Pd-HHDMA 1,2,7,9,10,30 Al Fe [23] hydrogenation catalysts based on: (i) alternative metals , (ii) Ce/TiO [27] 13 4 2 [24] 8,11–16,18–21 [28,29] Pd /Cu alloy and intermetallic compounds , (iii) ligand-modified PdB 1 6,17,22,25 26,27,34,35 Ag/SiO [30] [26] nanoparticles , (iv) transition metal oxides ,and(v) 2 CeO2 24,31,32 Pd@C N [31] Pd@C N supported single atoms . While these share better ensemble 3 4 [33] 3 4 [32] Pd4S/CNF definition than unmodified Pd, intrinsic drawbacks like instability Ag@C3N4 44–46 [34] against high pressure induced segregation for intermetallics , FeOx /ZrO2 fi Pd S/C N accessibility constraints in ligand-modi ed and single atom 3 3 4 In O [35] 31,47 48 2 3 catalysts , and poor H2 activation on oxides have been identified. In the search for new catalysts, very few works have studied the Pd S/C N modification with p-block elements22,28,29,49. In particular, sulfur- 3 3 4 modified Pd catalysts could be envisaged mimicking the strategy of ensemble and electronic density control in enzymes. In fact, Fig. 1 Survey of alkyne semi-hydrogenation catalysts. Chronological transition metal sulfides are applied for hydrotreating in oil development for gas- and liquid-phase applications tracing from the 1 2 6– refineries due to their resistance to poisons and ability toward H2 pioneering works of Wilde and Sabatier to the more recent discoveries activation50,51, and have also been studied for electrochemical 35. Six prominent catalyst families can be distinguished based on (i) 52 hydrogen evolution . Recently, a crystalline Pd4S phase was supported metal nanoparticles, (ii) intermetallics, alloys, and Lindlar-type shown to exhibit promising performance in the high pressure formulations where Pd is poisoned by a second metal, (iii) ligand-modified gas-phase hydrogenation of simple alkynes impurities in or hybrid materials featuring an organic surfactant that partially covers the -derived olefin streams, although no molecular-level surface of Pd nanoparticles, (iv) metal oxides, (v) supported single atoms, understanding was gathered33,53,54. Since the Pd-S phase diagram and (vi) Pd modified by p-block elements. The current industrially applied shows full mixing in all compositions and a continuum of closely catalysts are indicated in bold. Surface structures of representative related stable phases55, it offers room to tailor the active phase examples of each catalyst family are illustrated, highlighting the catalytic with the most selective ensemble and amount of electronic poi- ensemble soning at the atomic scale. By introducing a facile and scalable procedure to prepare alkynes. This is linked to the presence of trimeric metal ensem- crystal phase and orientation controlled Pd3S nanoparticles sup- fi ported on carbon nitride, this article demonstrates that sulfur can bles, which are also shown to be present in palladium sul des of play a multifunctional role enabling unparalleled hydrogenation different stoichiometry by the preparation and detailed inter- performance in the continuous-flow three-phase semi- rogation of the supported Pd4S phase. The superiority with hydrogenation of linear, internal, and bulky functionalized respect to other major families of hydrogenation catalysts is

2 NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4 ARTICLE a Pd Pd3S Table 1 Properties of the catalysts studied

Catalyst Pda/wt.% S b/ V c/ d d/nm D e/% Na S BET pore Pd CO 2 m2 g–1 cm3 g–1 T = 373–423 K Pd/C3N4 0.4 278 0.74 1.7 63 g Pd3S/C3N4- 0.4(1.9) 265 0.51 1.9 34 423f Pd4S/CNF 1.0 28 0.09 2.5 36 b PdPb/ 4.5 10 0.03 15 6 CaCO3 Pd-HHDMA 0.5 170 0.64 8.0 37 Pd@C3N4 0.5 155 0.26 0.4 —

a 0123 0 1 2 3 ICP-OES bBET method d / nm d / nm c Pd Pd Volume of N2 adsorbed at p/p0 = 0.95 dTEM eCO chemisorption fTemperature of sulfidation in K gMolar Pd/S ratio, the sulfur content was determined by elemental analysis

phases57, but this toxic and highly corrosive gas is synthetically c unattractive. Here, aqueous sodium sulfide was applied as an alternative, less widely studied sulfidation agent and the phase selectivity was monitored at different temperatures (Fig. 2a). Mesoporous graphitic carbon nitrite (C3N4) was targeted as a 0.250 nm carrier due to its well-known ability to stabilize palladium with 0.250 nm 31 high dispersion, including as single atoms (Pd@C3N4) , which was expected to facilitate the sulfidation treatment. Interestingly, despite exhibiting high stability in oxidative and reductive environments at elevated temperatures, tests in the temperature – range 373 423 K over Pd@C3N4 revealed that isolated palladium atoms were not stable against sulfidation treatments, transform- d ing into nanoparticles, as evidenced by scanning transmission 0.256 nm electron microscopy (STEM) (Supplementary Fig. 1a). For this reason, palladium was introduced via the wet deposition of pal- ladium chloride coupled with an additional sodium borohydride reduction step to promote the formation of palladium nano- 0246 8 – particles (Pd/C3N4) with an average diameter of 1 2 nm and a d / nm Pd very narrow size distribution. fi Analysis of the bulk properties of Pd/C3N4 and the sul ded analogs (Pd3S/C3N4-T, where Pd3S was the only crystalline phase identified and T = 373–473 K) confirmed the uptake of sulfur, absence of metal leaching, the negligible incorporation of sodium (< 2 ppm), and structural integrity of the carrier upon sulfidation Fig. 2 Synthetic approach and microscopy images of the palladium sulfide treatment (Table 1). The latter was verified by equivalent catalysts. a Graphical depiction of the sulfidation of palladium nanoparticles treatment of the metal-free C N establishing the preserved supported on graphitic carbon nitride. The average particle size increased 3 4 crystallinity and negligible sulfur uptake (Supplementary Fig. 2). from 1.2 to 1.8 nm upon treatment with Na S (see Supplementary Note 1), 2 Consistent with the low content and high dispersion of consistent with the distortion of the palladium lattice expected upon palladium, no metal-containing phases were detected by X-ray introduction of sulfur atoms. Scale bars: 30 nm. b HAADF-STEM image diffraction (XRD) in any of the catalysts supported on C N .To with the particle size distribution (inset) of Pd S/C N -423 fresh (left) and 3 4 3 3 4 determine whether the treatment resulted in the formation of one after 50 h on stream (right) excluding the occurrence of . Scale or more Pd S phases or sulfur doping localized at the surface, bars: 2 nm. c HRTEM image of Pd S/C N -423 fresh (left) and after 50 h x y 3 3 4 high-resolution transmission electron microscopy (HRTEM) on stream (right) confirming the preserved crystalline structure. d HAADF- analyses were undertaken. Examination of the HRTEM images STEM image (left) and the derived Pd particle size distribution (inset), and revealed the formation of crystalline Pd S nanoparticles, char- HRTEM (right) image of Pd S/CNF. Scale bars: 30 and 2 nm, respectively 3 4 acterized by an interplanar distance of 0.25 nm corresponding to the (202) facet56, in all of the sulfided samples on C N (Fig. 2c established by determining the hierarchy of key descriptors for 3 4 and Supplementary Fig. 3). Elemental mapping by energy- the performance. dispersive X-ray spectroscopy (EDX) further confirmed the close proximity of palladium and sulfur and absence of any sulfidation Results of the carrier (Supplementary Fig. 4). Synthesis of supported palladium sulfides. Sulfided palladium To gain insight into the thermodynamically preferred surface phases (PdxSy) can be readily prepared by the post-synthetic terminations of Pd3S, the stability of several stoichiometric low- modification of supported metal nanoparticles with sulfur- index planes was considered by density functional theory (DFT). containing compounds such as H2S, Na2S, PdSO4, and Pd The simulations indicate that Pd3S has two low-energy crystal thiolates. Several works have examined the phase evolution on facets: (001) and (202), both exhibiting inequivalent palladium 56 treatment of Pd/C with H2S , which often leads to mixed crystal and sulfur atoms (Fig. 3 and Supplementary Fig. 14). The Pd3S

NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4

Pd3S(202) Pd3S(001) seen as a lattice distortion to incorporate S anions, leading to a kinetically-controlled process where the formation of the (202) surface implies the lowest number of distortions. This agrees with earlier results identifying the preferential development of Pd3S (202) surface terminations upon sulfidation of carbon-supported palladium nanoparticles with hydrogen sulfide56. The chemical state of Pd at the surface of the sulfided catalysts was probed by X-ray photoelectron spectroscopy (XPS) (Supple- mentary Fig. 5). All the samples display Pd 3d peaks at 335.5 + 5/3 and 337.9 eV assigned to Pd0 and Pd2 species, respectively53. + Evaluation of the spectra shows that the relative ratio of Pd0/Pd2 increases upon sulfidation, pointing to the formation of sulfided Pd4S(200) Pd4S(110) Pd species with an average oxidation state close to 0. Note that there is still a debate on the XPS signals attributed to PdxSy species; some works reporting the presence of cationic Pd species58 and other observing a ‘metallic’ nature. The latter were attributed to the possible formation of a solid solution between Pd and S. To address this point and discern the bonding character, the difference charge density of Pd3S crystal and the charge transfer from Pd to S have been evaluated by the Bader charge analysis (Supplementary Note 3). The results point to a high degree of covalency, with a polar Pd–S bond and negligible electron transfer, which is in agreement with the experimentally observed tendency towards Pd0. In addition to charge effects, XPS fi Pd(111) Pd(110) shifts could also arise due to the nite size of the small nanoparticles in the catalyst that can also lead to deviations from the bulk values. To further explore the effect of sulfur poisoning on the catalysts electronic properties, projected d-band densities of states were calculated for all of the systems studied (Supplementary Fig. 15). Consistent with the XPS observations, – – down-shifts to 1.80 and 1.90 eV are evidenced for Pd3S(202) and Pd3S(001), respectively, compared to the reference value of – 1.39 eV for Pd(111). For reference purposes, a supported Pd4S catalyst was prepared fl via thermal decomposition in H2 ow of PdSO4 on carbon nanofibers (CNF)33. This procedure is less attractive since for

Fig. 3 Structural resemblance of Pd3S and Pd4S to pristine Pd surfaces. Top every mole of Pd4S produced it evolves three moles of toxic views of the low-energy terminations. The structural correspondence of the gaseous SO2 or even H2S. Furthermore, as previously reported, it sulfide surfaces with the most closely related palladium surface Pd(111) or lacks of a precise control over the particle size distribution (1–20 Pd(110) is highlighted with an orange hexagon or rectangle. The Pd atoms nm). Characterization by XRD confirms the formation of the on the Pd3S and Pd4S surfaces belonging to the ensemble are connected Pd4S phase (Supplementary Fig. 2a), and contrarily to Pd3S/C3N4, with a white triangle. Inequivalent palladium and sulfur atoms are identified the resulting crystals mainly expose the lowest-energy (200) facet with different colors (for more details see Supplementary Note 3). Color (Fig. 2d). Interestingly, a similar theoretical analysis of the Pd4S code: Pd(1)surf (purple), Pd(2)surf (blue), Pdbulk (light blue), S(1)surf (yellow), (200) surface reveals that it also features Pd trimers spaced by S S(2)surf (orange) atoms as well as a small fraction of Pd-square arrangements (Fig. 3). Consistent with the lower amount of sulfur, analysis by XPS evidences the closely metallic nature of surface palladium (001) surface displays Pd atoms arranged in chains and (peak at 335.5 eV, Supplementary Fig. 5) and a downshift of the − surrounded by S atoms, where rectangular patterns resembling Pd d-band to 1.70 eV, a slightly less poisoned state than Pd3S. those of the (110) surface of an fcc structure are found between the surface Pd neighbors. On the Pd3S(202) surface, each Pd is surrounded by four Pd atoms and one S. Therefore, both surfaces Performance in alkyne semi-hydrogenation. The sulfided pal- show triangular Pd ensembles with an area of 7.09 Å2 separated ladium catalysts were evaluated in the continuous-flow three- by S atoms. While on Pd3S(202), the triangular ensembles belong phase semi-hydrogenation of 2-methyl-3-butyn-2-ol, an impor- to the topmost layer, on Pd3S(001) they are formed by two tant building block in the synthesis of vitamin E, at various surface and one subsurface Pd atoms. The surface energy (γ )of temperatures and pressures (Fig. 4a, b and Supplementary Fig. 7) − s these terminations differs by 0.15 J m 2 with the (001) facet and benchmarked against state-of-the-art materials (Lindlar-type fi slightly favored (Supplementary Table 4). PdPb/CaCO3, ligand-modi ed Pd-HHDMA, Pd@C3N4 the Comparative inspection of the Pd3S(202) and Pd(111) surfaces properties of which are provided in Table 1 and Supplementary sheds light on the observed preferential development of the Table 1). Compared to the Lindlar catalyst, which achieves ~90% former, identifying their close geometrical relationship, main- selectivity to 2-methyl-3-buten-2-ol and the remaining 10% to 2- taining the same pseudo-hexagonal arrangement of the surface Pd methyl-2-butanol, Pd3S/C3N4-423 is fully selective to the (111) atoms in the fcc structure typically adopted by this metal. In and shows almost a 3.5-fold higher reaction rate (0.3 and 1.0 × 3 –1 contrast, the Pd3S(001) surface displays a closer resemblance to 10 h , respectively, at 303 K and 1 bar). Under the same con- the rectangular pattern of Pd(110). Therefore, starting from the ditions, Pd-HHDMA and Pd@C N catalysts are slightly less – 3 4 metallic Pd nanoparticles the mild sulfidation treatment can be active (0.6 and 0.2 × 103 h 1, respectively) although they display

4 NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4 ARTICLE

a Pd3S/C3N4-423 Pd4S/CNF Pd@C 3N4 PdPb/CaCO3 Pd-HHDMA 8

3.5 3.5 2.5 1.5 6 3.0 3.0

2.5 0.2 2.5 2.0 / bar

P 4 2.0 1.0 2.0 1.5 1.5 1.5 0.5 2 1.0 1.0 0.1

303 323 343 363 323 343 363 323 343 363 323 343 363 323 343 363 T /K

b 8

80 90 95 6 95 95 90 60

/ bar 4 70 P

80 70 2 90

303 323 343 363 323 343 363 323 343 363 323 343 363 323 343 363 T /K

Fig. 4 Evaluation of the alkyne semi-hydrogenation performance of the catalysts. a The contour plots map the effect of temperature and pressure on the 3 –1 rate (in 10 h ) and b selectivity to 2-methyl-3-buten-2-ol (in %) of Pd3S/C3N4-423 compared to Pd4S/CNF, PdPb/CaCO3, Pd-HHDMA, and Pd@C3N4. 3 –1 3 –1 Conditions: Wcat = 0.1 g, T = 303 K, FL(alkyne + toluene) = 1cm min , FG(H2) = 36 cm min

excellent selectivity (100%). Pd4S/CNF displays similar selectivity 80% conversion over the Pd3S and Pd4S phases, whereas much to Pd3S/C3N4-423 under all conditions applied, but with slightly more significant losses were evidenced for the PdPb/CaCO3 inferior reaction rates, for instance at 303 K and 1 bar the rate is (80%) and Pd-HHDMA (85%) catalysts. 3 –1 30% lower (0.7 compared to 1.0 × 10 h than on Pd3S/C3N4- To gain insight into the comparative reaction mechanisms over 423). Interestingly, Pd/C3N4 presents lower selectivity to the Pd and PdxS surfaces, DFT calculations were performed (details alkene (90%), but similar activity to Pd3S/C3N4-423, likely due to in Supplementary Note 4). Acetylene was chosen as a model the higher dispersion of palladium in the untreated sample alkyne to obtain robust answers to the challenging reaction (Table 1). networks. The validity of using this surrogate was verified by the To provide a broader comparison among all state-of-the-art parallel and reaction profiles observed for acetylene systems and extrapolate these conclusions, the reaction tempera- and 2-methyl-3-butyn-2-ol over the Pd3S catalysts (Supplemen- ture and pressure were varied to study the influence on the rate tary Fig. 17). Although the adsorption of 2-methyl-3-butyn-2-ol is and selectivity in the semi-hydrogenation of 1-hexyne, a widely slightly more exothermic, due to the hydrogen bond between the studied model alkyne, where the reaction can be accompanied by group of the alkynol and the surface sulfur atom, this the cis-trans isomerization of the alkene product (Supplementary interaction does not change the reaction profile as is present for Fig. 8). The contour maps clearly show the superiority of using p- all intermediates and products. Also, under liquid-phase opera- block elements for modifying Pd-based catalysts. Note that the tions the reaction environment has a limited effect on the state of only side product formed over the palladium sulfides was n- the catalyst59. Alkyne hydrogenation follows a Horiuti-Polanyi 59,60 – fi hexane (<5%). Unlike in the cases of Pd-HHDMA, PdPb/CaCO3, mechanism (Fig. 5, Supplementary Tables 5 7). The rst step and Pd@C3N4, no traces of isomers like cis and trans 2-hexene of the reaction is the dissociation of molecular hydrogen that were produced on PdxS catalysts. Furthermore, the similarly high occurs homolytically at surface Pd sites over Pd(111), Pd4S(110), 61 selectivity observed below 343 K and 5 bar, points to the Pd4S(200), and Pd3S(001), while the alternative heterolytic path β resistance of Pd3S/C3N4-423 towards the formation of -hydrides. features low energy only on the Pd3S(202) surface. This results in Consistently, analysis by the temperature-programmed reduction the storage of hydrogen atoms in the form of groups (SH) β 5 in H2 did not evidence any peak due to -hydrides . The DFT on the Pd3S(202) surface. C2H2 adsorption is exothermic on Pd – – – results indicate that the adsorption of H atoms in subsurface (111), Pd3S(202), and Pd4S(200) surfaces ( 2.51, 0.48, and 0.24 positions is endothermic by 0.14 eV. Since it was not possible to eV, respectively). Note that on Pd both the alkyne and H2 fl reach higher alkyne conversions in our continuous- ow set up, compete for the same site, while on Pd3S this is not the case additional batch tests were conducted to evaluate the selective because H atoms bind to surface sulfur atoms and the organic performance under more challenging conditions. An impressive reactant at the Pd-ensembles. This bifunctional mechanism, 95% selectivity to 2-methyl-3-buten-2-ol was preserved even at where fragments are stored at different sites of the catalyst

NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4

a 1 b –2 Pd3S(202)

Pd4S(200) –3

0 Pd(111) / eV E * * * Δ 3 3 2 –4 3 CCH CCH CCH CCH 3 2 2 3 H H H –5 H –1 c / eV E Δ –2 1 * / eV 4 H E 4 Δ C 2HCCH* 0

–3 2 * * 2 * 2 e 1 2 3 CCH CCH 2 2 2H* HCCH* HCCH + H HCCH H HCCH d –4 H 0 Pd(111) / eV E Δ –1 Pd4S(200)

Pd S(202) 3 –2

Pd(111) Pd4S(200) Pd3S(001) Pd3S(202)

Fig. 5 Density functional theory reaction analysis. a–c Energy profiles for the semi-hydrogenation (a), over-hydrogenation (b), and oligomerization (c)of acetylene on Pd3S(202), Pd4S(200), and Pd(111). The asterisks denote adsorbed species, the color of the profile matches that of the possible intermediate, and the bars relate to the energy differences for Pd3S(202) presented in e: HCCH adsorption (black bars), H2CCH2 desorption (gray bars), and the difference in activation energy between H2CCH2 (ethene) and HCCH3 (ethylidene) formation (red bars), H2CCH2 desorption and H2CCH3 formation (orange bars), and for oligomerization (blue bars). d Top view of the DFT-optimized adsorption configuration of the reaction intermediates and transition states (TS) on the surfaces. The elementary steps are shown in Supplementary Table 5. Color code as in Fig. 3, carbon (black), hydrogen (white)

surface, ensures higher activity of PdxS than the reference Pd/ Long-term tests conducted to verify the stability of the Pd3S/ fi Al2O3 (Supplementary Note 2). The rst addition of H to the C3N4-423 and Pd4S/CNF catalysts, evidenced no signs of activated alkyne leads to a vinyl moiety (HCCH2), whereas the deactivation after 50 and 25 h on stream, respectively (Fig. 6a). second can result in either ethene (H2CCH2) or ethylidene Unfortunately, a longer test on the Pd4S/CNF sample was not (HCCH3) formation. Notably, the difference in energy barrier possible due to the development of substantial pressure drop in between H2CCH2 and HCCH3 formation on Pd3S(202) is greater the reactor, which resulted from the mechanical instability of the than on Pd(111) and on Pd4S(200). At this stage, the desired CNF. Analysis post-reaction by HRTEM, STEM, and XPS fi semi-hydrogenation product (H2CCH2) could desorb or undergo con rmed the similar composition and absence of nanoparticle over-hydrogenation via an ethyl intermediate (H2CCH3) to form sintering or phase segregation with respect to the fresh catalyst for ethane. Ethene desorption from the Pd3S(202) and Pd4S(200) Pd3S/C3N4-423 (Fig. 2, Supplementary Fig. 5). This is an surfaces (0.08 and 0.22 eV, respectively) is energetically preferred important advantage with respect to intermetallic compounds, over its further hydrogenation (0.79 and 0.55 eV, respectively), where segregation effects have been reported to limit the even when the zero point vibrational energy corrections are ensemble control on the surface20,44,46,63,64. To shed light on included (Supplementary Table 6), which is completely opposite the intrinsic stability of the system, detailed simulations were to the behavior encountered on Pd(111) where the over- conducted to assess the likelihood of vacancy formation, sulfur hydrogenation barrier 0.45 eV is much lower than that of the incorporation to the reacting alkyne, and segregation effects. The alkene desorption (0.85 eV). This ultimately explains the low energy penalty for forming a vacancy and releasing H2S on the selectivity to ethene on Pd. The second potential side reaction (202) and (001) surfaces was calculated to be 0.56 and 1.58 eV, investigated is oligomer formation, which was studied by respectively, pointing to the stability under liquid-phase pro- modeling ethyne-ethyne coupling (Fig. 5c) and alkyne-vinyl cesses, while the formation energy of H2C = CHSH on the (202) paths providing similar insights due to scaling relations62 and (001) surfaces is 1.13 and 2.60 eV, highlighting the (Supplementary Fig. 19, 21). The barrier for oligomerization is impossibility of transferring sulfur to the substrate. Also, the 1.34 eV on the Pd(111) surface, and 1.64 eV on Pd3S(202), which segregation of a Pd atom from the bulk towards the surface indicates that the surface anisotropy hinders the diffusion of the requires 2.90, 4.30, and 2.96 eV for Pd3S(001), Pd3S(202), and alkyne on the latter surface, consistent with the expected site Pd4S(200), respectively, meaning that even in the presence of high isolation. This explains why almost no oligomer formation is amounts of atomic H on the surface no preferential Pd or S observed over the sulfided catalyst. Complementary analysis of segregation occurs44. Similarly, the formation of islands is not the reaction profiles on the Pd3S(001) and Pd4S(110) surfaces are favored as the energy to exchange a Pd by a surface S atom is 3.12, provided in Supplementary Note 4. 2.20, and 1.43 eV, respectively.

6 NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4 ARTICLE

a 2 a 1.2 –1 Pd3S h 3 1 / 10

r 0.8

Pd S Pd-HHDMA 0 4 –1

0 1020304050 h tos / h 3 PdZn

/ 10 0.4 b 3 r PdPb

Pd atoms Pd

OH Ag 0.0 Au 2 CeO2 –1 h 3 0.5 0.0 –0.5 –1.0 –1.5 –2.0 / 10 r Eads(C2H2) - Eads(C4H2) / eV 1

b 3 –1 OH r / 10 h 1.0

0 56789 0.8 E (C H ) / eV Eads(C2H4) / eV L / Å ads 2 2 ads 0.6 0.0 Fig. 6 Stability and substrate scope of the palladium sulfide catalysts. 0 a Both Pd3S/C3N4 (blue circles) and Pd4S/CNF (red triangles) display a 0.4 –0.5 stable performance on stream in the selective hydrogenation of 2-methyl- –1 L 3-butyn-2-ol. b The impact of the adsorption length ( ads) of functionalized Selectivity alkynes of different sizes over Pd3S/C3N4-423 (blue circles) and Pd- –2 HHDMA (green circles). Conditions: Wcat = 0.1 g, T = 303 K, P = 1(a)or 3 –1 3 −1 40 3(b) bar, FL(alkyne + toluene) = 1cm min , FG(H2) = 36 cm min Activity 1.5 30 1.0 0 20 10 Most of the relevant alkynes for the pharmaceutical and fine 0.5 1 chemical industries are highly branched compounds with internal 0.0 triple bonds and additional functionalities. To explore the scope 2 of supported Pd3S systems, the performance of Pd3S/C3N4-423 3 2 Ea(H2) / eV Aensemble / Å has been compared with that of the commercial ligand-modified 4 Pd-HHDMA catalyst in the hydrogenation of alkynes of varying size and functionality, namely, 1-hexyne, 3-hexyne, 1-dodecyne, 5 and 1,1-diphenyl-2-propyn-1-ol (Fig. 6b). With adsorption Eseg / eV fi lengths (de ned by the size of the activated alkyne) shorter than Stability 8 Å, similar rates were observed over both catalysts. However, while Pd-HHDMA becomes significantly less active at longer Pd S/C N Pd-HHDMA Pd/Al O adsorption lengths, the performance of Pd3S/C3N4-423 is 3 3 4 2 3 independent of the size and functionality of the substrate. This Pd3@C3N4 PdPb/CaCO3 PdZn/Al2O3 Au/TiO2 Ag/SiO2 effect can be rationalized by the fact that the architecture of active CeO2/TiO2 ensembles in Pd S is two dimensional as in the Lindlar system, 3 Pd4S/CNF whereas the HHDMA-modified catalyst shows a three dimen- sional structure featuring hydrophobic pores that provide access Fig. 7 Catalytic descriptors for alkyne semi-hydrogenation. a Rate of alkene to the alkynes only if they are terminal and not too bulky47. Thus, formation at T = 303 K and P = 1 bar versus the differential adsorption we can conclude that the Pd3S/C3N4-423 catalyst matches and energy of acetylene and ethene for all classes of catalysts. b Radar plot even exceeds the performance of Pd-HHDMA, illustrating the charting the relation of the rate with key theoretical indicators for the effectiveness of the sulfur modification for preparing superior activity: the adsorption energy of the alkyne (Eads(C2H2)) and the activation catalysts with clear practical scope. energy for H2 dissociation (Ea(H2)), selectivity: the alkene adsorption energy (Eads(C2H4)), and the ensemble area (Aensemble), and the stability: the segregation energy (Eseg). Values closest to the perimeter lead to Discussion enhanced properties. The blue shaded area connects the optimal values to

To overcome the intrinsic limitations of palladium nanoparticles date found for the Pd3S system (i.e., uncontrolled ensemble size and strong adsorption of the alkene product), researchers have mainly targeted the integration of confined ensembles (no more than 3 atoms) and tailored

NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4 electronic properties (Fig. 1). To compare those strategies with and 3.0 eV, respectively), while in PdPb this requires 0.7 eV. fi fi PdxS catalysts, we rst evaluated the alkyne semi-hydrogenation When considering ligand-modi ed systems like Pd-HHDMA, the performance using the differential adsorption energies, three dimensional structure of such catalysts imposes accessibility – fi Eads(C2H2) Eads(C2H4), as a uni ed descriptor accounting for constraints limiting their applicability for the conversion of bulky the intrinsic thermodynamic selectivity (Fig. 7a). Notice that and internal alkynes, important intermediates in the fine chemical other previously reported criteria for gas-phase hydrogenation industry. This limitation is not present in the PdxS catalyst as a screening such as the energy of adsorption38 or the two dimensional structure ensures the conversion of complex difference between the hydrogen barrier and the adsorption sterically hindered substrates. Single atoms display distinct geo- energy of the alkene65 do not provide a satisfactory description. metric and electronic properties compared to catalysts based on The former parameter does not consider the effects of van der nanoparticles, while the scaffold can also contribute to H2 acti- 31 Waals contributions, while the latter parameter is not suitable for vation and stabilization of the reactants . Sulfur atoms in PdxS catalysts in which hydrogen dissociation is rate determining (e.g., also share some of these features. The adsorbed H atoms are oxides). Even though all the modified materials display favorable stored at the S centers in the form of leaving the Pd site free values with respect to bare palladium (becoming more positive), for alkyne adsorption. This means that there is no competition correlation with the observed rates of 2-methyl-3-buten-2-ol between the alkyne and hydrogen for site availability and the formation reveals a wide variation in the productivity depending mechanism is similar to that of a dual site, which is again a on the catalyst family. Thus, a framework and hierarchy of consequence of the Pd–S bond polarity. The bifunctionality is property-performance relationships were identified to account more effective as no competition between the alkyne and not only for selectivity, but also activity and stability (Fig. 7b). hydrogen for the active sites occurs, this is reflected in the dif- Selectivity depends on (i) the adsorption energy of the alkene ferent reaction orders with respect to the alkyne, which is negative (Eads(C2H4)) which shows an optimal value close to 0 or positive, for Pd and close to zero for Pd3S (Supplementary Note 2). This 2 and (ii) the ensemble area whose improved value (ca. 8 Å )is explains the higher activity of Pd3S/C3N4 than the reference Pd/ shared by all catalysts. Activity is driven by two criterion with Al2O3. Therefore, the combination of crystal phase and orienta- fi different relative weight: the adsorption of the alkyne (described tion control observed in the sul dation process to form the Pd3S/ by the negative Eads(C2H2)) is favorable over all catalysts, while C3N4 catalyst ensures that the best values for the set of descriptors fi most importantly the hydrogen activation energy (Ea(H2)) dis- are achieved in an elegant, yet ef cient manner constituting a criminates between materials able of readily split molecular revolutionary step in catalyst design. hydrogen and catalysts requiring harsh reaction conditions. In conclusion, supported palladium sulfides have been shown Consideration of these factors clearly emphasizes the superiority to possess unique ensembles with face- and phase-dependent of Pd-based systems, exhibiting the lowest barriers for H2 dis- geometric and electronic characteristics and high stability, sociation and strong alkyne adsorption energies. Finally, stability demonstrating exciting potential as catalysts for the selective could be assessed by several potential phenomena and scenarios, hydrogenation of alkynes. A mild sulfidation treatment over but for the sake of simplicity the segregation energy (Eseg), which supported metallic Pd nanoparticles led to the incorporation of is the differential energy between the ground state (i.e., an alloyed sulfur into the palladium lattice, which drives the formation of a metallic surface) and the system after segregating one atom nanostructured Pd3S phase preferentially exposing the (202) towards the surface, is taken as a measure for the robustness of crystal facets. This surface displays isolated Pd3 ensembles sur- the catalyst. rounded by sulfur atoms featuring the right ensemble and elec- Compared to the archetypal Lindlar catalyst, sulfur in PdxS tronic poisoning to allow the selective hydrogenation of linear, systems would embed both the functions of acting as a molecular branched, and bulky functionalized alkynes. Ensembles of similar modulator on the ensemble mimicking the cooperative lead- architecture and functionality were also evidenced in Pd4S/CNF. 42 quinoline role, and the electronic poisoning effect of lead . The PdxS catalysts outperformed all state-of-the-art catalysts in However, sulfur poisoning is much more effective as very small terms of alkene formation rate per mole of Pd, and evidenced nanoparticles with a high surface/volume ratio can be used high durability with no sign of segregation. A detailed compar- instead of the large nanoparticles required for the PdPb catalyst. ison to the existing hydrogenation catalysts illustrated the hier- The need for facile hydrogen activation limits the practical rele- archy of key performance descriptors, demonstrating that the vance of oxides and other metals (Ag, Au)19,26,34,35,66. This is not reported synthetic protocol produces a catalyst that synergistically fi the case in PdxS, intermetallics, ligand-modi ed nanoparticles, combines the best characteristics of bond polarization, crystal and single-atom catalysts due to the intrinsic ability of Pd to phase and orientation control, while keeping a two-dimensional fi activate H2. The excellent site isolation in well-de ned ensembles nature enabling the superior performance for a broad set of featured by intermetallics, ligand-modified nanoparticles, and compounds. The present approach illustrates how the concept of single-atom catalysts lead to quite selective catalysts16,23,25,31. ensemble control in kinetically-trapped crystal orientations can However, the preferential hydrogen adsorption on Pd in alloys be employed in the precise design of evolved catalysts. and intermetallics induces severe segregation effects spoiling the control of surface ensembles achieved via the precise synthesis45,46. Note that induced segregation issues are much Methods Catalyst preparation. The mesoporous graphitic carbon nitride carrier (C3N4) was more relevant in gas-phase applications due to the high tem- prepared by controlled polymerization. Briefly, cyanamide (3.0 g, Sigma-Aldrich, perature, pressure, and absence of an organic solvent (Supple- 99%) and an aqueous dispersion of colloidal silica (12 g, 40 wt.% solids, 12 nm mentary Table 8). Remarkably, Pd segregation or islanding by average diameter, Sigma-Aldrich) were mixed and stirred at 373 K for 6 h to completely evaporate the water. The resultant solids were ground and calcined at substituting a palladium atom with sulfur is not favorable due to – – 823 K (2.3 K min 1 ramp) for 4 h under a nitrogen flow (15 cm3 min 1). The silica the polarization of the Pd-S bond enhancing the robustness of 3 was removed by treating the solid in a 4 M aqueous solution of NH4HF2 (200 cm , Pd3S. This is even more enhanced in Pd3S(001) and Pd4S(110) as Acros Organics, 95%) for 48 h. The template free mesoporous carbon nitride was fi both display vacancy formation energy higher than Pd3S(202), collected by ltration, washed thoroughly with distilled water and ethanol, and dried at 333 K overnight. making Pd4S more suited to high-pressure gas-phase operation. The incorporation of palladium nanoparticles was approached by wet That is the reason why among all catalysts the Eseg (accounting deposition, targeting a metal loading of 0.5 wt.%. The carrier was dispersed in –3 for the energy needed to enlarge the surface ensemble by one Pd deionized water (0.01 gcarrier cm ) under sonication for 1 h prior to the 3 –1 atom from the bulk) of Pd3S and Pd4S are the highest ones (5.0 introduction of an aqueous solution (4.72 cm gcarrier ) containing PdCl2 (0.01 M,

8 NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4 ARTICLE

ABCR Chemicals, 99%) and NaCl (0.04 M, Sigma-Aldrich, 99%). The resulting 99%) as internal standard, and toluene (Fischer Chemicals, 99.95%). The reactions suspension was stirred at room temperature for 2 h in the dark. To promote metal – – – were conducted at various temperatures (303 363 K) and total pressures (1 8 bar), nanoparticle formation, an aqueous solution (1.4 cm3 g 1) of NaBH (0.5 M, 3 –1 3 −1 fl carrier 4 keeping constant liquid (1 cm min ) and H2 (36 cm min ) ow rates. The Sigma-Aldrich, 99%) was subsequently added dropwise and the mixture was stirred reaction orders were obtained as a function of the gas-phase H pressure and the 2- fi 2 overnight. Finally, the solids (Pd/C3N4) were collected by ltration, washed methyl-3-butyn-2-ol concentration. Blank alkyne semi-hydrogenation experiments thoroughly with distilled water and ethanol and subsequently dried at 333 K over the range of conditions investigated, excluded the competitive hydrogenation overnight. The atomic dispersion of palladium on C3N4 (Pd@C3N4) was achieved of or toluene. The stability test was carried out at T = 303 K and P = 1 bar. by microwave-assisted-deposition. An aqueous solution of Pd(NH3)4(NO3)2 (0.05 The products were collected after 20 min of steady-state operation. The batch 3 3 cm , 5 wt.% Pd, ABCR Chemicals) was added to a suspension (20 cm ) of the hydrogenation was carried out in a CEM Discover SP microwave reactor with a dispersed C3N4 (0.5 g) and stirred overnight in the dark. The resulting suspension pressure-controlled vessel under continuous stirring. In a typical reaction, the was treated in a microwave reactor (CEM Discover SP), applying a cyclic program liquid feed (1.5 cm3) was heated in the presence of the catalyst (0.01–0.02 g) at 323 of 15 s irradiation and 3 min cooling with 20 repetitions using a power of 100 W. K. The initial H2 pressure was 3 bar. In all cases, the products were analyzed offline The resulting powder was collected, washed, and dried as described for the using a gas chromatograph (HP-6890) equipped with a HP-5 capillary column and nanoparticle-based system. a flame ionization detector. The conversion (X) of the substrate was determined as fl The catalyst was added to a Te on-lined autoclave containing an aqueous the amount of reacted substrate divided by the initial amount of substrate. The solution of Na2S·9H2O (Sigma-Aldrich, 99%) using a molar S/Pd ratio of 10. The reaction rate (r) was expressed as the number of moles of 2-methyl-3-buten-2-ol autoclave was then heated to the desired temperatures (373, 398, or 423 K) for 3 h. produced per hour and total moles of Pd. The selectivity (S) to each product was fi The solid was collected by ltration, extensively washed with deionized water and quantified as the amount of the particular product divided by the amount of ethanol, and dried overnight at 333 K. reacted substrate. The carbon balance exceeded 98% in all cases, excluding the In an alternative route, a supported Pd4S catalyst was prepared by incipient formation of oligomers. Consistently, negligible differences were observed between wetness impregnation of PdSO4 (ABCR-Chemicals, 99.99%) on CNFs (Aldrich). the thermogravimetric profiles of both the fresh and used catalyst. After drying the resulting material was treated under 10 vol.% H2/He at 523 K for 1h. Computational details. The reaction network for hydrogenation on Pd and PdxS Reference PdPb/CaCO3 (Alfa Aesar, ref: 043172) and Pd-HHDMA, where surfaces was investigated by means of DFT calculations with periodic boundary HHDMA stays for hexadecyl(2-hydroxyethyl)dimethylammonium dihydrogen 67 phosphate (NanoSelect LF 200TM, Strem Chemicals, ref.46-1711) catalysts were conditions within the RPBE approach with a plane wave cut-off energy of 450 eV (for more details see Supplementary Method 1 and Note 4) and dispersion con- used as received. No quinoline was added to the PdPb/CaCO3 catalyst during tributions with D2 parameters68,69 complemented with our reparametrized values testing. 70 for metals . The models for the Pd3S and Pd4S surfaces were constructed with six S-Pd-S trilayers where the two topmost S-Pd-S trilayers were fully relaxed, whereas Characterization methods. The Pd and Na contents in the catalyst were deter- the four bottommost layers were fixed to their bulk positions. The bare Pd(111) mined by inductively coupled plasma-optical emission spectrometry (ICP-OES) surface was modeled as a periodically repeated p(3 × 2) slab consisting of five using a Horiba Ultra 2 instrument equipped with a photomultiplier tube detector. atomic layers, where the three at the bottom were fixed and the two on the top were The samples were dissolved in piranha solution under sonication until the absence fully relaxed. Due to the operation in liquid phase and the low binding energies of visible solids. The S content was determined by infrared spectroscopy using a found for the reactants, no significant deviation from the original stoichiometry is LECO CHN-900 combustion furnace. Nitrogen isotherms were obtained at 77 K observed, which differs from gas-phase operation59. The climbing image-modified using a Micrometrics Tristar instrument, after evacuation of the samples at 423 K nudged elastic band (CI-NEB) method71,72 was employed to locate the transition for 3 h. The Brunauer–Emmett–Teller (BET) method was applied to calculate the states. To date, the solvent contributions to the evaluation of Gibbs energies is still fi total surface area (SBET). High-resolution transmission electron microscopy under debate, thus, the energy pro les are presented in terms of the potential (HRTEM), high-angle annular dark field STEM (HAADF-STEM), and elemental energies. The contributions of zero-point vibrational energy (Supplementary mapping with EDX spectroscopy were conducted on a FEI Talos microscope Table 6) and vibrational entropies to competing steps are similar for bifurcation operated at 200 kV. Samples were prepared by dusting respective powders onto steps, thus their differences are small enough to be safely neglected. The computed lacey-carbon coated grids. The particle size distribution was assessed by structures can be retrieved from the ioChem BD database73 via the link (https://doi. counting more than 100 individual Pd nanoparticles. CO chemisorption was org/10.19061/iochem-bd-1-57). performed using a Micromeritics Autochem II 2920 chemisorption analyzed coupled with a MKS Cirrus 2 quadrupole mass spectrometer. The sample was – Data availability. The authors declare that the data supporting the findings of this pretreated in situ at 393 K under He flow (20 cm3 min 1) for 60 min, and reduced fi fl 3 –1 3 study are available within the article and its Supplementary Information le. All at 348 K under owing 5 vol.% H2/N2 (20 cm min ) for 30 min. Thus, 0.491 cm other relevant source data are available from the corresponding author upon of 1 vol.% CO/He was pulsed over the catalyst bed at 308 K every 4 min. The reasonable request. palladium dispersion was calculated from the amount of chemisorbed CO, con- – sidering an atomic surface density of 1.26 × 1019 atoms m 2 and an adsorption stoichiometry Pd/CO = 2. Powder XRD was measured using a PANalytical X’Pert Received: 4 December 2017 Accepted: 7 June 2018 PRO-MPD diffractometer and Cu-Kα radiation (λ = 0.154 nm). The data was recorded in the 10–70° 2θ range with an angular step size of 0.017 and a counting time of 0.26 s per step. XPS was conducted with a Physical Electronics Instruments Quantum 2000 spectrometer using monochromatic Al Kα radiation generated from an electron beam operated at 15 kV and 32.3 W. The spectra were recorded References − under ultra-high vacuum conditions (residual pressure = 5×10 8 Pa) at a pass 1. Wilde, M. Pv vermischte mittheilungen. Ber. Dtsch. Chem. Ges. 7, 352 (1874). energy of 50 eV. In order to compensate for charging effects, all binding energies 2. Sabatier, P. & Senderens, J. B. Action of hydrogen on acetylene in presence of were referenced to the C 1 s at 288.2 eV. 13C solid-state cross-polarization/magic nickel. C. R. Chim. 128, 1173 (1899). angle spinning nuclear magnetic resonance (CP/MAS NMR) spectra were recorded 3. Borodziński, A. & Bond, G. C. Selective hydrogenation of ethyne in ethene‐ on a Bruker AVANCE III HD NMR spectrometer at a magnetic field of 16.4 T rich streams on palladium catalysts. Part 1. Effect of changes to the catalyst 1 corresponding to a H Larmor frequency of 700.13 MHz. A 4 mm double reso- during reaction. Catal. Rev. 48,91–144 (2006). nance probe head at a spinning speed of 10 kHz was used for all experiments. The 4. Eggersdorfer, M. et al. One hundred years of vitamins - a success story of the 13 C spectra were acquired using a cross polarization experiment with a contact natural sciences. Angew. Chem. Int. Ed. 51, 12960–12990 (2012). 3 time of 2 ms and a recycle delay of 1 s. A total of 64 × 10 scans were added for each 5. McCue, A. J. & Anderson, J. A. Recent advances in selective acetylene 13 1 sample. C experiments used high-power H decoupling during acquisition using hydrogenation using palladium containing catalysts. Front. Chem. Sci. Eng. 9, a SPINAL-64 sequence. 142–153 (2015). 6. Paal, C. & Hohenegger, C. Über katalytische Wirkungen kolloidaler Metalle Catalytic testing der Platingruppe. XII. Die stufenweise Reduktion des Acetylens. Ber. Dtsch. . The hydrogenation of 2-methyl-3-butyn-2-ol (Acros Organics, – 99.9%), 1-hexyne (Acros Organics, 98%), 3-hexyne (Acros Organics, 99%), 1- Chem. Ges. 48, 275 287 (1915). dodecyne (Acros Organics, 98%), 4-octyne (Acros Organics, 98%), and 1,1- 7. Bolton, E. K. & Downing, F. B. Process of preparing butadiene. US patent diphenyl-2-propyn-1-ol (Sigma-Aldrich, 99 %) was carried out in a fully- 1,777,600 (1930). ’ automated flooded-bed reactor (ThalesNano H-Cube ProTM), in which the gaseous 8. Campbell, K. N. & O Connor, M. J. The hydrogenation of substituted – hydrogen, produced in situ by the electrolysis of Millipore water, and the liquid acetylenes with . J. Am. Chem. Soc. 61, 2897 2900 (1939). feed flow concurrently upward through a cartridge of approximately 3.5 mm 9. Sauerwein, K. Production of from acetylene. US patent 1,860,624 internal diameter. The latter contains a fixed bed which is composed of the catalyst (1932). (0.10 g) well mixed with carbide as diluent (0.12 g, Aldrich, 99.8%), both 10. Soday, F. J. Process for refining unsatuated hydrocarbon material with an with particle size of 0.2–0.4 mm. The potential role of heat- and mass-transfer alkali or alkaline earth metal. US patent 2,413,254 (1946). limitations was excluded using the Koros-Nowak diagnostic test (Supplementary 11. Lindlar, H. Ein neuer Katalysator für selektive Hydrierungen. Helv. Chim. Note 2). The liquid feed contained 5 wt.% of substrate, 3 wt.% of benzene (Merck, Acta 35, 446–450 (1952).

NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4

12. Frevel, L. K. & Kressley, L. J. Selective hydrogenation of acetylene in ethylene 43. Sheldon, R. A. & van Bekkum, H. (eds) in Fine Chemicals through and catalyst therefor. US patent 2,802,889 (1957). 351–471 (Wiley-VCH Verlag GmbH, Weinheim, 13. Leeds, M. W. & Tedeschi, R. J. Process for the production of olefinic carbinols. 2007). US patent 2,989,567 (1961). 44. López, N. & Vargas-Fuentes, C. Promoters in the hydrogenation of alkynes in 14. Jean, B. G. Palladium-tin catalysts. US patent 3,192,168 (1965). mixtures: insights from density functional theory. Chem. Commun. 48, 15. Dubuis, R. & Lindlar, H. Process for selective hydrogenation. US patent 1379–1391 (2012). 3,715,404 (1973). 45. Sankar, M. et al. Designing bimetallic catalysts for a green and sustainable 16. Sarkany, A., Zsoldos, Z., Furlong, B., Hightower, J. W. & Guczi, L. future. Chem. Soc. Rev. 41, 8099–8139 (2012). Hydrogenation of 1-butene and 1,3-butadiene mixtures over Pd/ZnO 46. Rameshan, C. et al. In situ XPS study of methanol reforming on PdGa near- catalysts. J. Catal. 141, 566–582 (1993). surface intermetallic phases. J. Catal. 290, 126–137 (2012). 17. Mastalir, Á. & Király, Z. Pd nanoparticles in hydrotalcite: mild and highly 47. Vilé, G., Almora-Barrios, N., Mitchell, S., López, N. & Pérez-Ramírez, J. From selective catalysts for alkyne semihydrogenation. J. Catal. 220, 372–381 (2003). the Lindlar catalyst to supported ligand-modified palladium nanoparticles: 18. Choudhary, T. V., Sivadinarayana, C., Datye, A. K., Kumar, D. & Goodman, selectivity patterns and accessibility constraints in the continuous-flow three- D. W. Acetylene hydrogenation on Au-based catalysts. Catal. Lett. 86,1–8 phase hydrogenation of acetylenic compounds. Chem. Eur. J. 20, 5926–5937 (2003). (2014). 19. Bridier, B. & Pérez-Ramírez, J. Cooperative effects in ternary Cu-Ni-Fe 48. García-Melchor, M., Bellarosa, L. & López, N. Unique reaction path in catalysts lead to enhanced alkene selectivity in alkyne hydrogenation. J. Am. heterogeneous catalysis: the concerted semi-hydrogenation of propyne to – – Chem. Soc. 132, 4321 4327 (2010). propene on CeO2. ACS Catal. 4, 4015 4020 (2014). 20. Osswald, J. et al. Palladium-gallium intermetallic compounds for the selective 49. Pinna, F. et al. Consecutive hydrogenation of benzaldehyde over Pd catalysts: hydrogenation of acetylene. J. Catal. 258, 210–218 (2008). influence of supports and sulfur poisoning. Appl. Catal. A 219, 195–200 21. Semagina, N., Grasemann, M., Xanthopoulos, N., Renken, A. & Kiwi-Minsker, (2001). L. Structured catalyst of Pd/ZnO on sintered metal fibers for 2-methyl-3- 50. Chianelli, R. R. et al. Catalytic properties of single layers of transition metal butyn-2-ol selective hydrogenation. J. Catal. 251, 213–222 (2007). sulfide catalytic materials. Catal. Rev. 48,1–41 (2006). 22. McKenna, F.-M. & Anderson, J. A. Selectivity enhancement in acetylene 51. McCue, A. J. & Anderson, J. A. Sulfur as a catalyst promoter or selectivity fi fi – hydrogenation over diphenyl sulphide-modi ed Pd/TiO2 catalysts. J. Catal. modi er in heterogeneous catalysis. Catal. Sci. Technol. 4, 272 294 (2014). – 281, 231 240 (2011). 52. Hinnemann, B. et al. Biomimetic hydrogen evolution: MoS2 nanoparticles as – 23. Armbrüster, M. et al. Al13Fe4 as a low-cost alternative for palladium in catalyst for hydrogen evolution. J. Am. Chem. Soc. 127, 5308 5309 (2005). heterogeneous hydrogenation. Nat. Mater. 11, 690–693 (2012). 53. Bachiller-Baeza, B. et al. Detecting the genesis of a high-performance carbon- 24. Kyriakou, G. et al. Isolated metal atom geometries as a strategy for selective supported Pd sulfide nanophase and its evolution in the hydrogenation of heterogeneous . Science 335, 1209–1212 (2012). butadiene. ACS Catal. 5, 5235–5241 (2015). 25. Witte, P. T. et al. BASF NanoSelect™ technology: innovative supported Pd- 54. McCue, A. J., Guerrero-Ruiz, A., Ramirez-Barria, C., Rodríguez-Ramos, I. & and Pt-based catalysts for selective hydrogenation reactions. Top. Catal. 55, Anderson, J. A. Selective hydrogenation of mixed alkyne/alkene streams at 505–511 (2012). elevated pressure over a palladium sulfide catalyst. J. Catal. 355,40–52 26. Vilé, G., Bridier, B., Wichert, J. & Pérez-Ramírez, J. Ceria in hydrogenation (2017). catalysis: high selectivity in the conversion of alkynes to olefins. Angew. Chem. 55. Zubkov, A., Fujino, T., Sato, N. & Yamada, K. Enthalpies of formation of the Int. Ed. 51, 8620–8623 (2012). palladium sulphides. J. Chem. Thermodyn. 30, 571–581 (1998). fi 27. Vilé, G. et al. Stereo- and chemoselective character of supported CeO2 56. Xu, W. et al. Tailoring supported palladium sul de catalysts through H2- fl fi – catalysts for continuous- ow three-phase alkyne hydrogenation. assisted sul dation with H2S. J. Mater. Chem. A 1, 12811 12817 (2013). ChemCatChem 6, 1928–1934 (2014). 57. Vázquez-Zavala, A., Fuentes, S. & Pedraza, F. The influence of sulfidation on 28. Chan, C. W. A. et al. Interstitial modification of palladium nanoparticles with the crystalline structure of palladium, rhodium and catalysts boron atoms as a green catalyst for selective hydrogenation. Nat. Commun. 5, supported on silica. Appl. Surf. Sci. 78, 211–218 (1994). 5787 (2014). 58. Singh, V. V., Kumar, U., Tripathi, S. N. & Singh, A. K. Shape dependent fl 29. Yang, B., Burch, R., Hardacre, C., Hu, P. & Hughes, P. Selective hydrogenation catalytic activity of nano owers and nanospheres of Pd4S generated via one of acetylene over Pd–boron catalysts: a density functional theory study. J. pot synthesis and grafted on graphene oxide for Suzuki coupling. Dalton Phys. Chem. C 118, 3664–3671 (2014). Trans. 43, 12555–12563 (2014). 30. Vilé, G. & Pérez-Ramírez, J. Beyond the use of modifiers in selective alkyne 59. García-Mota, M., Bridier, B., Pérez-Ramírez, J. & López, N. Interplay between hydrogenation: silver and gold nanocatalysts in flow mode for sustainable carbon monoxide, hydrides, and carbides in selective alkyne hydrogenation on alkene production. Nanoscale 6, 13476–13482 (2014). palladium. J. Catal. 273,92–102 (2010). 31. Vilé, G. et al. A stable single-site palladium catalyst for hydrogenations. 60. Horiuti, I. & Polanyi, M. Exchange reactions of hydrogen on metallic catalysts. Angew. Chem. Int. Ed. 54, 11265–11269 (2015). Trans. Faraday Soc. 30, 1164–1172 (1934). 32. Chen, Z. et al. Merging single-atom-dispersed silver and carbon nitride to a 61. García-Melchor, M. & López, N. Homolytic products from heterolytic paths in joint electronic system via copolymerization with silver tricyanomethanide. H2 dissociation on metal oxides: the example of CeO2. J. Phys. Chem. C 118, ACS Nano 10, 3166–3175 (2016). 10921–10926 (2014). 33. McCue, A. J., Guerrero-Ruiz, A., Rodríguez-Ramos, I. & Anderson, J. A. 62. Vignola, E. et al. Evaluating the risk of C–C bond formation during selective Palladium sulphide - a highly selective catalyst for the gas phase hydrogenation of acetylene on palladium. ACS Catal. 8, 1662–1671 (2018). hydrogenation of alkynes to . J. Catal. 340,10–16 (2016). 63. Coq, B. & Figueras, F. Bimetallic palladium catalysts: influence of the co-metal 34. Tejeda-Serrano, M. et al. Synthesis of supported planar iron oxide on the catalyst performance. J. Mol. Catal. A Chem. 173, 117–134 (2001). nanoparticles and their chemo- and stereoselectivity for hydrogenation of 64. Wowsnick, G. et al. Surface dynamics of the intermetallic catalyst Pd2Ga, Part alkynes. ACS Catal. 7, 3721–3729 (2017). II - Reactivity and stability in liquid-phase hydrogenation of phenylacetylene. 35. Albani, D. et al. Semihydrogenation of acetylene on indium oxide: proposed J. Catal. 309, 221–230 (2014). single-ensemble catalysis. Angew. Chem. Int. Ed. 56, 10755–10760 (2017). 65. Yang, B., Burch, R., Hardacre, C., Headdock, G. & Hu, P. Origin of the 36. Bridier, B., López, N. & Pérez-Ramírez, J. Molecular understanding of alkyne increase of activity and selectivity of nickel doped by Au, Ag, and Cu for hydrogenation for the design of selective catalysts. Dalton. Trans. 39, acetylene hydrogenation. ACS Catal. 2, 1027–1032 (2012). 8412–8419 (2010). 66. Hammer, B. & Nørskov, J. K. Why gold is the noblest of all the metals. Nature 37. Crespo-Quesada, M., Cárdenas-Lizana, F., Dessimoz, A.-L. & Kiwi-Minsker, 376, 238–240 (2009). L. Modern trends in catalyst and process design for alkyne hydrogenations. 67. Makov, G. & Payne, M. C. Periodic boundary conditions in ab initio ACS Catal. 2, 1773–1786 (2012). calculations. Phys. Rev. B 51, 4014–4022 (1995). 38. Studt, F. et al. Identification of non-precious metal alloy catalysts for selective 68. Grimme, S. Semiempirical GGA-type density functional constructed with a hydrogenation of acetylene. Science 320, 1320–1322 (2008). long-range dispersion correction. J. Comput. Chem. 27, 1787–1799 (2006). 39. Vilé, G., Albani, D., Almora-Barrios, N., López, N. & Pérez-Ramírez, J. 69. Bučko, T., Hafner, J., Lebègue, S. & Ángyán, J. G. Improved description of the Advances in the design of nanostructured catalysts for selective structure of molecular and layered crystals: ab initio DFT calculations with hydrogenation. ChemCatChem 8,21–33 (2016). van der Waals corrections. J. Phys. Chem. A 114, 11814–11824 (2010). 40. Cybulski, A., Moulijn, J. A., Sharma, M. M. & Sheldon, R. A. (eds) in Fine 70. Almora-Barrios, N., Carchini, G., Błoński, P. & López, N. Costless derivation Chemicals Manufacture 59–124 (Elsevier Science B.V., Amsterdam, 2001). of dispersion coefficients for metal surfaces. J. Chem. Theory Comput. 10, 41. Pollak, P. (ed.) in Fine Chemicals: The Industry and the Business,8–25 (John 5002–5009 (2014). Wiley & Sons, Inc., Hoboken, 2011). 71. Henkelman, G., Uberuaga, B. P. & Jónsson, H. A climbing image nudged 42. García-Mota, M. et al. A density functional theory study of the ‘mythic’ elastic band method for finding saddle points and minimum energy paths. J. Lindlar hydrogenation catalyst. Theor. Chem. Acc. 128, 663–673 (2011). Chem. Phys. 113, 9901–9904 (2000).

10 NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05052-4 ARTICLE

72. Henkelman, G. & Jónsson, H. Improved tangent estimate in the nudged elastic Competing interests: The authors declare no competing interests. band method for finding minimum energy paths and saddle points. J. Chem. Phys. 113, 9978–9985 (2000). Reprints and permission information is available online at http://npg.nature.com/ 73. Álvarez-Moreno, M. et al. Managing the computational chemistry big data reprintsandpermissions/ problem: the ioChem-BD platform. J. Chem. Inf. Model. 55,99–103 (2015). Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Acknowledgements Financial support from ETH Zurich, the Swiss National Science Foundation (Grant No. 200021-169679), and MINECO (grant number CTQ2015-68770-R). The Barcelona Supercomputing Centre (BSC-RES) and ScopeM of ETH Zurich are thanked for access to Open Access This article is licensed under a Creative Commons their facilities. Dr René Verel is thanked for help with the MAS NMR measurements. M. Attribution 4.0 International License, which permits use, sharing, S. thanks the COFUND program and MINECO for support through Severo Ochoa adaptation, distribution and reproduction in any medium or format, as long as you give Excellence Accreditation 2014-2018 (SEV-2013-0319). appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party Author contributions material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the J.P.-R. and N.L. conceived and coordinated all stages of this research. D.A. and Z.C. article’s Creative Commons license and your intended use is not permitted by statutory prepared and characterized the samples. D.A. performed the catalytic tests. S.M. acquired the microscopy images. R.H. undertook the XPS analyses. M.S. conducted the DFT regulation or exceeds the permitted use, you will need to obtain permission directly from simulations. All authors contributed to writing the manuscript. the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. Additional information Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- © The Author(s) 2018 018-05052-4.

NATURE COMMUNICATIONS | (2018) 9:2634 | DOI: 10.1038/s41467-018-05052-4 | www.nature.com/naturecommunications 11